How To Come Up With Good Predictor Variables

[aha123]

I would like to know how you create good predictor variables for a regression model.

Here is a scenario:

Suppose a web site would like to build a regression model to predict the conversion rate for each product listed in the web site. We have two price information. First is the web site's price for each product(PRICE). Second is the lowest price among all competitors(LOWEST_PRICE). There are many other variables which can be used as predictor variables. But let's focus on PRICE and LOWEST_PRICE here. I want to know if you will create additional predictor variables based on these 2. If so, how?

[Reeza]

Right off the bat, calculate the difference and the percent difference.

price - lowest_price = price difference

price difference/price = percent price

[aha123]

Is there any guideline for creating derived predictor variables? If you know any articles or papers on this, please share.